Processing apparatus, processing method, and program

Abstract

This processing device 1 comprises: a storage device that stores definition data 12 specifying a false positive function for detecting, as change point candidates in time-series data, all times serving as change points, and detecting times other than the change points, and a false negative function for detecting, as change point candidates in time-series data, times of a portion of the change points among the change points, and not detecting times other than the change points; and an output unit 23 that outputs, as change points, times detected by the false negative function as the change point candidates from time-series data 11 to be processed, among times detected by the false positive function as the change point candidates from the time-series data 11 to be processed.

Description

Technical Field

[0001] The present invention relates to a processing device, a processing method, and a program.

Background Art

[0002] There is a demand to appropriately detect change points in time-series data. For example, in the monitoring of network devices, there are various requirements for the change points to be detected, such as detecting a change in the increasing trend of network devices or detecting an increase in periodic amplitude.

[0003] There is a method of customizing an analysis method according to each requirement of the change point to be detected. For example, regarding the monitoring of network devices, when monitoring a network device by performing a balance analysis using multiple types of time-series data related to the network device, there is a method of detecting abnormalities in the network device with high accuracy (for example, Patent Document 1). In Patent Document 1, among multiple types of time-series data, multiple types of time-series data whose degree of relationship is within the range of allowable error are targeted for balance analysis.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, the method of customizing the analysis method according to each requirement of the change point to be detected has low versatility and may require high development costs.

[0006] The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a technique capable of efficiently detecting change points in time-series data.

Means for Solving the Problems

[0007] A processing device according to one aspect of the present invention includes a false positive function that detects all times that constitute change points as change point candidates in time series data, as well as times other than change points; a storage device that stores definition data for identifying a false negative function that detects the times of some of the change points as change point candidates in time series data, and does not detect times other than the change points; and an output unit that outputs as change points the times that the false negative function detects as change point candidates from the time series data to be processed, among the times that the false positive function detects as change point candidates from the time series data to be processed.

[0008] A processing method according to one aspect of the present invention involves a computer storing definition data in a storage device that identifies a false positive function which detects all times that constitute change points as candidate change points in time series data, as well as times other than change points, and a false negative function which detects the times of some of the change points as candidate change points in time series data, and does not detect times other than the change points. The computer then outputs as change points the times that the false negative function detects as candidate change points from the time series data to be processed, among the times that the false positive function detects as candidate change points from the time series data to be processed.

[0009] One aspect of the present invention is a program that makes a computer function as the processing device. [Effects of the Invention]

[0010] According to the present invention, it is possible to provide a technology that can efficiently detect change points in time-series data. [Brief explanation of the drawing]

[0011] [Figure 1] Figure 1 is a diagram illustrating the functional block of the processing apparatus according to an embodiment of the present invention. [Figure 2] Figure 2 illustrates the false positive and false negative functions. [Figure 3]Figure 3 illustrates the output sets of the false positive and false negative functions. [Figure 4] Figure 4 illustrates the data structure of the definition data and an example of the data. [Figure 5] Figure 5 is a flowchart illustrating the output processing performed by the output unit. [Figure 6] Figure 6 is a diagram illustrating the hardware configuration of the computer used in the processing unit. [Modes for carrying out the invention]

[0012] Embodiments of the present invention will be described below with reference to the drawings. In the drawings, identical parts are denoted by the same reference numerals and their descriptions are omitted.

[0013] The processing apparatus 1 according to the embodiment of the present invention shown in Figure 1 outputs change points in the time-series data 11 to be processed.

[0014] The processing unit 22 of the processing unit 1 has processing units that process time series data 11 with each of a plurality of evaluation functions F1, F2, F3... and output calculated values ​​for each time. Each evaluation function F1, F2, F3... is a simple function that outputs a calculated value, which is a numerical representation of a value corresponding to a certain time in the time series data 11 obtained through a predetermined process, in association with that time. Examples of evaluation functions include a function to calculate whether or not there is periodicity in the time series data 11 and the period if there is periodicity, a function to calculate the rate of increase or decrease of values ​​over a predetermined period, a function to calculate the maximum or minimum value of values ​​over a predetermined period, a function to calculate the average value of values ​​over a predetermined period, and a function to calculate the amount or rate of change of values ​​over a predetermined period. In embodiments of the present invention, any evaluation function among the plurality of evaluation functions F1, F2, F3... may be denoted as evaluation function F.

[0015] The output unit 23 of the processing device 1 calculates the time-series data 11 using each of a plurality of evaluation functions F, and extracts change-point candidates from the calculated values using a plurality of functions. The output unit 23 combines the change-point candidates extracted by each of the plurality of functions according to the characteristics of the function, and identifies a change point from the change-point candidates. The output unit 23 outputs the time at the identified change point.

[0016] Generally, if one function tries to detect change points without excess or deficiency, the function may become complex. In contrast, the processing device 1 combines the change-point candidates extracted by each of the plurality of functions according to the characteristics of the function, and identifies a change point from the change-point candidates. The processing device 1 can use simple functions, and high versatility and suppression of development costs can be expected, so change points can be detected efficiently.

[0017] In an embodiment of the present invention, a false positive function and a false negative function are defined as functions for detecting change-point candidates.

[0018] The false positive function detects all times that become change points as change-point candidates in the time-series data, and also detects times other than change points. The false positive function is a function for detecting, from each time, at least all times that become change points, based on the calculated value of each time output by the evaluation function F. The false positive function allows for the detection of times that do not become change points. Since the false positive function widely detects change-point candidates in the time-series data, it is suitable for screening change-point candidates.

[0019] The false negative function detects the times of some of the change points as change-point candidates in the time-series data, and does not detect times other than change points. The false negative function is a function for detecting only a part of the times that become change points, from each time, based on the calculated value of each time output by the evaluation function F. The false negative function does not detect times that do not become change points. Since the times detected as change-point candidates by the false negative function become change points, it is suitable for detecting change points from the change-point candidates detected by the false positive function.

[0020] The processing device 1 can efficiently detect change points by outputting the change point candidates detected by the false negative function as change points among the change point candidates detected by the false positive function.

[0021] Fig. 2 shows the relationship between the detection points detected by the false positive function and the false negative function as change point candidates, and the change points output by the processing device 1 for certain time series data. In Fig. 2, the black symbols are the detection points detected by the false positive function or the false negative function, and the white symbols are the non-detection points not detected by the false negative function. The round symbols are the change points output by the processing device 1, and the triangular symbols are the non-change points not output by the processing device 1 as change points.

[0022] Referring to Fig. 2(a), the change point candidates detected by the false positive function will be described. The three symbols attached to the time series data shown in Fig. 2(a) are all represented in black and have a round shape and a triangular shape. The change point candidates detected by the false positive function include the change points to be detected and the non-change points that do not necessarily need to be detected as change points.

[0023] Referring to Fig. 2(b), the change point candidates output by the false negative function will be described. The three symbols attached to the time series data shown in Fig. 2(b) are all represented in a round shape and have black and white. The change point candidates detected by the false negative function include only some of the change points to be detected and do not include some of the change points to be detected.

[0024] Referring to Fig. 3, the inclusion relationship among the output set D output by the processing device 1 as change points, the output set FP output by the false positive function as change point candidates, and the output set FN output by the false negative function as change point candidates will be described. The output set FP of the false positive function includes the output set D. The output set FN of the false negative function is included in the output set D.

[0025] In embodiments of the present invention, the false positive function and the false negative function are identified, respectively, by a threshold value calculated by the evaluation function F and a determination condition that specifies whether the value is greater than or less than the threshold. The false positive function and the false negative function are set, for example, using test time series data. Desired change points are set in advance for the test time series data by an operator or the like. Here, the desired change points are points in the test time series data that the processing unit 1 is expected to detect as change points.

[0026] The threshold and criteria for identifying a false positive function are determined such that the time intervals of all predetermined change points, as calculated by the evaluation function F for the test time series data, are included, as well as the time intervals that are not change points. The threshold and criteria for identifying a false negative function are determined such that the time intervals of only some of the predetermined change points, as calculated by the evaluation function F for the test time series data, are included, and the time intervals that are not change points are excluded. Note that the false positive function and the false negative function may be identified using different evaluation functions, or they may be identified using the same function.

[0027] The false positive and false negative functions used by the processing unit 1 can be easily created because they can be identified from the test time series data, the change points in that test time series data, and the evaluation function F. Furthermore, since the processing unit 1 detects change points by combining the false positive and false negative functions, a sophisticated evaluation function is not required, and a simple evaluation function can be used. In this way, the processing unit 1 can easily detect change points using the false positive and false negative functions.

[0028] Furthermore, even if the requirements for the conditions to be detected as change points change, the processing unit 1 can appropriately detect the desired change points from the combination of outputs of false positive and false negative functions created to meet those requirements. Because such a processing unit 1 can efficiently detect change points even if the conditions to be detected as change points change, it is highly versatile and can reduce development costs.

[0029] The false positive and false negative functions are generated from desired change points set by the operator for the test time series data. It is possible that the false positive and false negative functions may not be able to properly detect candidate change points if the trends of the time series data 11 being processed differ from those of the test time series data, or if events not considered by the test time series data occur in the time series data 11. However, the processing unit 1 is expected to detect change points with greater accuracy by detecting change points using a combination of multiple functions with different perspectives.

[0030] (processing device) As shown in Figure 1, the processing unit 1 includes time-series data 11, definition data 12, and change point data 13, as well as the functions of an acquisition unit 21, a processing unit 22, and an output unit 23. Each piece of data is stored in a storage device such as memory 902 or storage 903. Each function is implemented in the CPU 901.

[0031] The time-series data 11 is data that associates time with a value corresponding to that time, such as monitoring data from network equipment. The time-series data 11 is the data that the processing unit 1 processes to detect change points. The time-series data 11 is referenced by the processing unit 22.

[0032] The definition data 12 identifies the false positive function and the false negative function, as shown in Figure 4. The definition data 12 is pre-set and referenced by the processing unit 22 and the output unit 23.

[0033] Definition data 12 identifies the range of calculated values ​​for candidate change points output by the false positive function, which is the calculated value output by a certain evaluation function. Definition data 12 also identifies the range of calculated values ​​for candidate change points output by the false negative function, which is the calculated value output by a certain evaluation function.

[0034] In the example shown in Figure 4, f1, f2, f3, and f4 represent the sets of calculated values ​​output by the evaluation functions F1, F2, F3, and F4, respectively. ">500", "<7.5", etc., are the threshold values ​​and judgment conditions for the calculated values ​​that become candidate change points within the set of calculated values ​​of the evaluation function. For example, the function "f1>500", set as a false positive function, indicates that it outputs times corresponding to calculated values ​​greater than 500 within the set of calculated values ​​f1 obtained from the time series data 11 and evaluation function F1 as candidate change points for the false positive function. The function "f3<7.5", set as a false negative function, indicates that it outputs times corresponding to calculated values ​​less than 7.5 within the set of calculated values ​​f3 obtained from the time series data 11 and evaluation function F3 as candidate change points for the false negative function.

[0035] The change point data 13 is data that identifies the change point output by the processing unit 1. The change point data 13 includes the time determined to be a change point in the time series data 11. The change point data 13 is output by the output unit 23.

[0036] The acquisition unit 21 acquires time-series data 11, which is the target of change point detection, from, for example, a communication system 2.

[0037] The processing unit 22 inputs the time series data 11 into each evaluation function F. The processing unit 22 inputs the time series data 11 into each evaluation function that outputs a set of calculated values ​​referenced in the definition data 12 to identify a false positive function or a false negative function. The definition data 12 shown in Figure 2 identifies a false positive function or a false negative function using sets of calculated values ​​f1, f2, f3, and f4. The processing unit 22 inputs the time series data 11 into evaluation functions F1, F2, F3, and F4, respectively.

[0038] The processing unit 22 inputs the set of calculated values ​​that each evaluation function F has calculated for each time point in the time series data 11, associating each calculated value with the corresponding time point, into the output unit 23. In the case of the definition data 12 shown in Figure 2, the processing unit 22 inputs the set of calculated values ​​f1 calculated by evaluation function F1 and the time points corresponding to each of those calculated values, the set of calculated values ​​f2 calculated by evaluation function F2 and the time points corresponding to each of those calculated values, the set of calculated values ​​f3 calculated by evaluation function F3 and the time points corresponding to each of those calculated values, and the set of calculated values ​​f4 calculated by evaluation function F4 and the time points corresponding to each of those calculated values ​​into the output unit 23.

[0039] The processing unit 22 may have a separate processing unit for each evaluation function F that it can process. As shown in Figure 1, the processing unit 22 includes a first processing unit 22a that processes evaluation function F1, a second processing unit 22b that processes evaluation function F2, a third processing unit 22c, and so on, that processes evaluation function F3. Each processing unit 22a, 22b, 22c, ... may process the time series data 11 in parallel or serially.

[0040] The output unit 23 outputs the time periods in which the false negative function detected as a candidate change point from the time series data 11 to be processed, out of the time periods in which the false positive function detected as a candidate change point from the time series data 11 to be processed, as change points.

[0041] The output unit 23 identifies candidate change points detected by the false positive function and the false negative function, respectively, by referring to the definition data 12 from the set of calculated values ​​input from the processing unit 22. The output unit 23 identifies the intersection of the candidate change points detected by the false positive function and the candidate change points detected by the false negative function as the change point. The output unit 23 outputs the time at the identified change point.

[0042] Definition data 12 defines one or more false positive functions and one or more false negative functions. Here, the processing of output unit 23 is explained for each combination where there is one or two (or more) false positive and false negative functions. In any combination, output unit 23 outputs the time of the change point candidate obtained by the intersection of output set FP and output set FN as the change point time.

[0043] (1) In the definition data 12, we will explain the case in which one function is set as the false positive function and one function is set as the false negative function.

[0044] When definition data 12 defines a function as a false positive function, the output set FP of the false positive function is the set of candidate change point times output by that function. The case where the function "f1>500" is set as the false positive function is explained below. The output unit 23 calculates the set of times fp1 corresponding to the calculated values ​​greater than 500 from the set of calculated values ​​f1 output by the evaluation function F1. The output unit 23 uses the calculated set fp1 as the output set FP of candidate change point times identified by the false positive function.

[0045] When definition data 12 defines a function as a false negative function, the output set FN of the false negative function is the set of time points of candidate change points output by that function. Let's explain the case where the function "f3 < 7.5" is set as the false negative function. The output unit 23 calculates the set of time points fn3 that corresponds to calculated values ​​less than 7.5 from the set of calculated values ​​f3 output by the evaluation function F3. The output unit 23 sets the calculated set fn3 as the output set FN of time points of candidate change points identified by the false negative function.

[0046] (2) This section describes the case where the definition data 12 identifies two functions as false positive functions and two functions as false negative functions.

[0047] When definition data 12 defines two functions as false positive functions, the output set FP of the false positive functions is the intersection of the sets of change point candidate times output by each of the multiple functions. Let's explain the case where the functions "f1>500" and "f2>20" are set as false positive functions in definition data 12. The output unit 23 calculates the set of times fp1 corresponding to calculated values ​​greater than 500 from the set of calculated values ​​f1 output by the evaluation function F1. The output unit 23 calculates the set of times fp2 corresponding to calculated values ​​greater than 20 from the set of calculated values ​​f2 output by the evaluation function F2. The output unit 23 takes the intersection of the calculated sets fp1 and fp2 as the output set FP of change point candidate times identified by the false positive functions. The output set FP of the false positive functions identified by two functions has fewer change point candidates other than the output set D compared to the set of false negative functions identified by one function, and the change point candidates can be narrowed down more effectively.

[0048] When definition data 12 defines two functions as false negative functions, the output set FN of the false negative functions is the union of the sets of change point candidate times output by each of the functions. Let's explain the case where the functions “f3<7.5” and “f4>120” are set as false negative functions in definition data 12. The output unit 23 calculates the set of times fn3 corresponding to calculated values ​​less than 7.5 from the set of calculated values ​​f3 output by the evaluation function F3. The output unit 23 calculates the set of times fn4 corresponding to calculated values ​​greater than 120 from the set of calculated values ​​f4 output by the evaluation function F4. The output unit 23 takes the union of the calculated sets fn3 and fn4 as the output set FN of change point candidate times identified by the false negative functions. The output set FN of false negative functions identified by two functions includes more change points in the output set D as change point candidates compared to the set of false negative functions identified by one function, so it can detect more change point candidates.

[0049] (3) The case in which the definition data 12 identifies two (or more) functions as false positive functions and one function as a false negative function is explained below. The output unit 23 outputs as change points the times at which the false negative function detects a change point candidate from the time series data 11 to be processed, among the times at which both of the two false positive functions detect a change point candidate. When the definition data 12 defines two functions as false positive functions, the output set FP of the false positive functions is the intersection of the sets of change point candidate times output by each of the multiple functions. When the definition data 12 defines one function as a false negative function, the output set FN of the false negative function is the set of change point candidate times output by that function.

[0050] (4) The case in which the definition data 12 identifies one function as a false positive function and two (or more) functions as false negative functions is explained below. The output unit 23 outputs as change points the times in which any of the multiple false negative functions have detected as change point candidates from the time series data 11 to be processed, among the times in which the false positive function has detected as change point candidates. When the definition data 12 defines one function as a false positive function, the output set FP of the false positive function is the set of change point candidate times output by that function. When the definition data 12 defines two functions as false negative functions, the output set FN of the false negative functions is the union of the sets of change point candidate times output by each of the multiple functions.

[0051] The output processing by the output unit 23 will be explained with reference to Figure 5. The content and sequence of processing shown in Figure 5 are examples only and are not limited thereto.

[0052] In step S1, the output unit 23 inputs the time-series data 11 to be processed from the processing unit 22 into each evaluation function and obtains the calculated value for each time.

[0053] For each time point in the time-series data 11 to be processed, the processing in steps S2 to S7 is performed. The output unit 23 determines whether the calculated value of the time to be processed satisfies each function defined in the definition data 12, and from the result, determines whether or not to output the time to be processed as a change point. Here, the first false positive function defined in the definition data 12 is "f1>500", and the second function is "f2>20". The third false negative function is "f3<7.5", and the fourth function is "f4>120".

[0054] In step S2, the output unit 23 makes a determination regarding the first function of the false positive function. If the calculated value of the processing time of the first evaluation function F1 is greater than 500, the process proceeds to step S3; otherwise, the process proceeds to step S7.

[0055] In step S3, the output unit 23 makes a determination regarding the second function of the false positive function. If the calculated value of the processing time for the second evaluation function F2 is greater than 20, the process proceeds to step S4; otherwise, the process proceeds to step S7.

[0056] In step S4, the output unit 23 makes a determination regarding the third function of the false negative function. If the calculated value of the processing time of the third evaluation function F3 is less than 7.5, the process proceeds to step S6; if it is 7.5 or greater, the process proceeds to step S5.

[0057] In step S5, the output unit 23 determines the fourth function of the false negative function. If the calculated value of the processing time of the fourth evaluation function F4 is greater than 120, the process proceeds to step S6; otherwise, the process proceeds to step S7.

[0058] In step S6, the time to be processed is output as a change point. Specifically, if both false positive functions determine that the time to be processed is a change point candidate, and either of the two false negative functions determines that it is a change point candidate, then the time to be processed is a change point.

[0059] In step S7, the time being processed is not output as a change point. Specifically, if, for the time being processed, one of the two false positive functions determines that it is not a candidate change point, and both of the two false negative functions determine that it is not a candidate change point, then the time being processed is not a change point.

[0060] The processing device 1 according to an embodiment of the present invention efficiently extracts change points by combining the change point candidates extracted by each of several functions according to the characteristics of each function, thereby identifying change points from the change point candidates. Furthermore, since change points are identified using multiple functions, it is possible to use simpler functions compared to identifying change points with a single function, and high versatility and reduced development costs can be expected.

[0061] Furthermore, by providing definition data 12 for each user and an output unit 23 for each user, the processing unit 1 may detect change points in the time-series data 11 of multiple users. The processing unit 1 can also detect change points that each user independently desires from each of the time-series data 11 of multiple users, according to the definition data 12 that each user has independently set. In this case, since each user can use the processing unit 22 in common, the cost of introducing new users can also be suppressed.

[0062] (modified version) The embodiments of the present invention described a case in which false positive and false negative functions are identified by setting threshold values ​​for the calculated values ​​calculated by the evaluation function F and the criteria for determining those threshold values, so as to satisfy the definitions of false positive and false negative functions. In contrast, the modified example describes a case in which the false positive function is identified from the relationship between change points set for the test time series data and the output results of the function.

[0063] This section explains how to generate a false positive function using a perfectly mismatched function. Here, a perfectly mismatched function is defined as a function that detects only the time points other than change points in time series data. In the example shown in Figure 3, the output set NCZ of the perfectly mismatched function contains only points that are not change points and does not overlap at all with the output set D of change points.

[0064] The false positive function detects all time points that constitute change points, as well as time points that are not change points. Therefore, the false positive function can be defined as a function that detects the complement of the set of time points output by the completely mismatched function. The output unit 23 may detect candidate change points of the false positive function according to this definition.

[0065] This section explains how to generate a false positive function using a partially mismatched function. Here, a partially mismatched function is defined as a function that detects the times of some of the change points in time-series data, as well as the times of points that are not change points. In the example shown in Figure 3, the output set NCZ of the partially mismatched function overlaps with part of the output set D of the change points, and also includes points that are not change points.

[0066] The false positive function detects all time points that constitute change points, as well as time points that are not change points. Therefore, if the union of the time points of each change point detected by each of the multiple partially non-conforming functions contains the output set D of change points, the false positive function can be defined as a function that detects the union of the sets detected by each of those multiple partially non-conforming functions. The output unit 23 may detect candidate change points of the false positive function according to this definition.

[0067] The processing unit 1 may output candidate change points for the false positive function using the false positive function identified by the definition in the modified example.

[0068] The processing unit 1 of this embodiment described above uses, for example, a general-purpose computer system comprising a CPU (Central Processing Unit, processor) 901, memory 902, storage 903 (HDD: Hard Disk Drive, SSD: Solid State Drive), communication device 904, input device 905, and output device 906. In this computer system, each function of the processing unit 1 is realized by the CPU 901 executing a program loaded onto the memory 902.

[0069] The processing unit 1 may be implemented on a single computer, or on multiple computers. Furthermore, the processing unit 1 may be a virtual machine implemented on a computer.

[0070] The program of the processing unit 1 can be stored on a computer-readable recording medium such as an HDD, SSD, USB (Universal Serial Bus) memory, CD (Compact Disc), or DVD (Digital Versatile Disc), or it can be distributed over a network. A computer-readable recording medium is, for example, a non-transitory recording medium.

[0071] It should be noted that the present invention is not limited to the embodiments described above, and numerous modifications are possible within the scope of its essence. [Explanation of Symbols]

[0072] 1 Processing Unit 2. Communication System 11. Time series data 12 Definition Data 13 Change Point Data 21 Acquisition Department 22 Processing Units 23 Output section 901 CPU 902 memory 903 Storage 904 Communication equipment 905 Input device 906 Output device

Claims

1. A memory device that stores definition data for a false positive function that detects all times that constitute change points as candidate change points in time series data, as well as times other than change points, and a false negative function that detects the times of some of the change points as candidate change points in time series data, and does not detect times other than the change points. Output unit outputs as change points the time periods in which the false positive function detects change point candidates from the time series data to be processed, and the time periods in which the false negative function detects change point candidates from the time series data to be processed. Equipped with, If the aforementioned definition data identifies multiple functions as false negative functions, The output unit is a processing device that outputs as change points the time at which any of the multiple functions of the false negative function detects a change point candidate from the time series data to be processed, among the time at which the false positive function detects a change point candidate.

2. If the aforementioned definition data identifies multiple functions as false positive functions, The output unit outputs the time period in which the false negative function detects a change point candidate from the time series data to be processed, among the times in which any of the multiple functions of the false positive function detect a change point candidate. The apparatus according to claim 1.

3. If the function detects only times other than change points in the time series data, the false positive function detects the complement of the set of times output by the function. The apparatus according to claim 1.

4. Each of the multiple functions detects the time of some of the change points in the time series data, and also detects the time of other change points. If the union of the time sets of each of the aforementioned functions that detects each change point includes the set of change points, the false positive function detects the union of the sets that each of the aforementioned functions detects. The apparatus according to claim 1.

5. The computer stores definition data in a memory device that identifies a false positive function that detects all time points that constitute change points as candidate change points in time series data, as well as time points other than change points, and a false negative function that detects the time points of some of the change points as candidate change points in time series data, and does not detect time points other than the change points. The computer outputs as change points the time periods in which the false positive function detects change point candidates from the time series data to be processed, and the time periods in which the false negative function detects change point candidates from the time series data to be processed. If the aforementioned definition data identifies multiple functions as false negative functions, The computer outputs as the change point any of the multiple functions of the false negative function that it detects as a change point candidate from the time series data to be processed, within the time periods that the false positive function detects as change point candidates. Processing method.

6. A program for causing a computer to function as a processing device according to any one of claims 1 to 4.

Images